- From: Andy Seaborne <andy.seaborne@epimorphics.com>
- Date: Tue, 10 Apr 2012 18:08:06 +0100
- To: Ivan Herman <ivan@w3.org>
- CC: rdf-wg <public-rdf-wg@w3.org>
On 10/04/12 15:35, Ivan Herman wrote: >> "Subgraphs or Graphs" >>> >>> I don't think this is a local syntax issue. The parts of the >>> labelling may be in different datasets, which are then brought >>> together (actually, not possible by condition 3) >>> > Yep, another reason to remove this. > > But I am not absolutely sure what you are saying:-( > >>> Within one TriG file, several<u> {} blocks may make one graph >>> labelled<u> overall -- that is a syntax issue. >>> >>> ((I don't see why the merge of datasets isn't the merge of their >>> graphs with the same URI + the (IRI, graph) pairs for >>> non-overlapping IRIs.)) >>> > Well, that may be doable... but this seems to be one (maybe the only > real?) open issue at the moment in the WG. > > Just thinking out loud: if (<u>,G) but<u> is_not_ of type > rdf:Graph, ie, it is only labeling, then I could imagine a much more > lax attitude in terms of subgraph vs. graphs. However, if<u> > rdf:type rdf:Graph, ie, it is really a URI that denotes the graph, > then the situation may be different... I'm OK with the idea that additional statements about <u> can, in effect, close the description by placing further restrictions on the relationship of <u> and G. I haven't seen a reason to make the default be complete labelling - additional triples can't undo the base semantics. e.g. Lee's example or Arnaud's "it depends" or simply concatenating two N-Quads or TriG files are reasonable UCs to me. Andy
Received on Tuesday, 10 April 2012 17:08:42 UTC